一种面向空间对象群的聚类算法

A Clustering Algorithm for Spatial Object Groups

将空间对象按一定的空间关系分组,构成空间对象群,每个空间对象群包含类型多样、数量不等的空间对象。提出一种空间对象群的相异度计算方法,并提出空间对象群聚类算法SOGC。它将类型多样的空间数据集分层表示,计算空间对象群中对象在不同层上属性分布的隶属度,以此计算空间对象群的相异度。与一般的聚类算法不同,SOGC考虑了空间数据的复杂性和数据之间的联系。实验结果表明算法SOGC是有效的,对地理空间数据分析具有实用价值。

There are spatial relationship between objects in the spatial database. The objects are divided into several groups by spatial relationship, which are named spatial object groups. There are numerous objects and many spatial data types in every spatial object group. The clustering algorithm for spatial ob- ject groups has seldom been studied. A computing method of dissimilarity between spatial object groups is proposed and a algorithm（named SOGC） is presented for clustering spatial object groups. The spatial objects are partitioned into layer by their types. The subjection of attribute distribution of the objects in spatial object groups is computed and the dissimilarity between spatial object groups is calculated based on the subjection. Algorithm SOGC takes variety data types and spatial relationships between spatial objects into full account, so it is different from the other clustering algorithms. The experimental results show that algorithm SOGC is effective for spatial data analysis.